Point-to-point MIMO and massive MIMO techniques have played significantroles in the success of 4G and 5G radio networks, and in 6G we believethat distributed MIMO will play a similar critical role. The performanceof downlink phase coherent distributed MIMO transmission relies on tightphase alignment between the serving access points (APs) in the system.In realistic scenarios, there will always be some level of phasemisalignment between APs due to e.g., differences in the local clocksof the APs, which can severely degrade the performance.

One main contribution of this paper is that we propose the use of aLinear Quadratic Regulator (LQR) based solution for calculating downlinkprecoding weights in D-MIMO systems.  The optimal LQR based precodingsolution is numerically stable and computationally efficient, and it caneasily utilise parallel computing in distributed or centralised hardwareprocessors.  Furthermore, we also show how the LQR based solution can bemodified to include differently sized subsets of serving APs for each UE,which enables a scalable tradeoff between performance and complexity.

Another main contribution of the paper is that we identify a new phasemisalignment problem in D-MIMO. The proposed LQR-based precoding methodis the first solution that takes not only the channel estimation phaseerrors, but also the relative phase errors between serving APs intoaccount when designing the downlink D-MIMO transmission precoder. By this,some of the performance lost due to different causes of phase misalignmentcan be regained. In the scenarios studied in this paper we observe 20-70%performance increase of the proposed method compared to a reference casewhere residual phase errors are ignored when determining the downlinkprecoding weights.